-
Notifications
You must be signed in to change notification settings - Fork 41
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Correct indexing of cospOUT #86
base: master
Are you sure you want to change the base?
Conversation
…sat_bugfix Pull updates to CI tests.
…sat_bugfix Merge from CFMIP master COSP2.0.
Hi @alejandrobodas, I think that I merged in the latest changes. But it looks like the integration failed again. If I am doing something incorrectly just let me know. Happy to continue pulling in from the latest changes. |
@jshaw35 I think we should expect some answer changes here. |
Hi @jshaw35 . Yes, the tests will fail because your changes modify the results. This makes the review of this change a bit complex, despite the simplicity of the code changes. We will guide you through the process. Basically, this is done in two steps:
Please can you add to this PR plots of the diagnostics that have changed? In order to do this, you'll need to download and unpack the outputs generated by one of the compilers, e.g. outputs.gfortran-12.UKMO.tgz. You'll find the files at the bottom of the failed tests: The files contain plots of the outputs that have been generated automatically by the integration tests, but unfortunately they don't contain plots of the variables that are affected by this PR, so you'll need to produce the plots offline. It is sufficient to add plots from the outputs of gfortran-12 only.
I hope this helps. Please let us know if you need more information and thanks for contributing to COSP! |
Hi @alejandrobodas, I've finally gotten around to producing the plots of the diagnostics that changed. I updated plot_test_outputs.py, so plots of the missing variables should now be produced during CI (here I've assumed some bounds on the CFODD_NDBZE and CFODD_NICOD axes). The updated plots after the recent failed CI are here: I don't actually use the diagnostic fields that were affected by the indexing error, so I will leave interpretation of the updated plots to you. Just reviewing them, however, I noticed that the histograms for the ncfodd1,2,3 variables almost always have in-cloud optical depth values in the 0-2 bin. Best, |
Sorry for the late response. I missed the PR notification. |
This PR has been stalled for a while. The code changes look fine, but as @jshaw35 mentions above the plots of the fields affected by the indexing don't look right to me. @takmichibata please can you have a look at the plots attached? |
Thank you for sharing the plots @alejandrobodas @jshaw35. |
Hi all,
The cospOUT DDT seems to be missing correct indexing for when inputs are chunked in the CloudSat quickbeam_column call (lines 1272-1277) and the cosp_diag_warmrain call (lines 1640-1670). I have corrected the indexing errors.
@dustinswales mentioned that there might be an issue in how the MODIS diagnostics are passed to the warm rain diagnostics. I think that adding the (ij:ik) indexing when passing cospOUT fields to cosp_diag_warm as inputs should align the other warm rain diagnostic inputs with cospOUT and fix things.
A simple test using the COSP offline driver showed changes relative to the KGO values when chunking is used. This is expected. There should be no changes when there is no chunking in calls to COSP_SIMULATOR.
Best,
Jonah